Florian Golemo

Since @BostonDynamics didn't answer our emails, we had to build our own Spot Mini #robot (just a bit more... shoddy 😁). Thanks to @MassCaccia, @chrisjpal, and @element_ai for helping out, @stanfordrobotx for the quad design and @poppy_project for the arm design.

🚗💥Introducing Ctrl-Crash: controllable video generation for autonomous driving! SOTA models struggle to generate physically realistic car crashes. We propose an image2video diffusion model with bounding box and crash type control. Website: anthonygosselin.github.io/Ctrl-Crash-Pro… 🧵->

Ctrl-Crash Controllable Diffusion for Realistic Car Crashes

1/2 Here's our story from last night about Elon Musk...youtube.com/watch?v=Eo3zOR…

me & my collaborators have done the most extensive research of the LAION datasets (3 academic papers & the first to investigate dataset in 2021 showing misogyny, pornography, & malignant stereotypes) yet, the Stanford study has not cited us once. this is academic misconduct

@DavidSKrueger Are there? What high-profile people in the AI research community think that?

There are a significant number of people in the AI research community who explicitly think humans should be replaced by AI as the natural next step in evolution, and the sooner the better!

The skill of being able to imitate others from just minimal visual observation is challenging to learn but an extremely beneficial skill most intelligent animals possess. We develop an agent that can learn to imitate an expert from only a single video demonstration—a 🧵.

Omfg @EriolDoesDesign thank you sooooo much! You're the best! 🤩

6. Apple Orange

@tyrell_turing Eyyyyyy huge congrats!!! Well deserved! 🎉 🎉

I've just received tenure 🎉 - just in time for a summer of AI existential risk conversations in highly public forums, which works well I guess... 🙃

@tyrell_turing 100%! IMHO this is the most salient and pressing point - there's AI used all over the place right now from recommendation algorithms that lead to increased polarization to credit assessment with untraceable judgement. This needs regulation more urgently than some hypothetical.

3/4) Moreover, we risk overlooking real immediate threats from AI (e.g. use of AI for manipulating populations, biased models being used to make decisions, etc.). And, we risk impairing research into beneficial applications of AI with this sort of hyperbole.

1/4) Several people I admire immensely have signed this, but respectfully, I'm afraid I just don't agree with the claim that "mitigating the risk of extinction from AI should be a global priority". I think this statement is naive and a mistake.

The problem isn't regulating "AI" or future "AGI". It's protecting individuals from corporate and government overreach using "AI" to cut costs and or deflect accountability.

And then some people go around saying we should accelerate this. 😬 I'm all for better AI assistants but let's take it easy and test them properly before mass deployment.

Announcing our newest preprint Features from CLIP, DINO are ready for zero-shot 3D perception. 𝗖𝗼𝗻𝗰𝗲𝗽𝘁𝗙𝘂𝘀𝗶𝗼𝗻 builds open-set multimodal 3D maps by fusing features to 3D. These maps can be queried by text, image, click, and audio📜👇 concept-fusion.github.io

The #2023DLRL Summer School will be held in #Montréal, presented by CIFAR and Mila, in collaboration with Amii and the Vector Institute. Submit your application for the #DLRLSchool before February 22! 👉 pulse.ly/zqh6cx5fxm

@LegalizeBrain @kareem_carr It's clearly a diagram, not a neuron.

I don’t know who needs to hear this but this is not a neuron. 😬

RL training has high variance over final policy performance. Our new work @corl_conf #Agilerobotics workshop augments RL by learning to estimate where the agent is underperforming and forcing the agent to learn along this boundary. W/ @velythyl, @shoddy_robots, @duckietown_coo

@irinarish Could you explain that, please? Because that's the contentious and important part. Is it okay to discount other people's well-being in the near-term for humanity's long-term well-being?

@shoddy_robots To some extent

"DESPITE EVERYTHING, THIS ISN’T A TIME TO GIVE UP ON EFFECTIVE ALTRUISM" vox.com/future-perfect…

@irinarish Does your version of longtermism entail discounting any current or near-term well-being in favor of long-term?

@shoddy_robots It means considering long-term implications of our current actions and planning for the well-being of humanity and intelligent beings very far in the future - definitely much longer term than the usual pressing problems - see What We Owe The Future by MacAskill, for example.

@irinarish What do you think "longtermism" means? In your own words?

@shoddy_robots And, to an extent, I am a longtermist, and I think everyone should be, too - of course, the question is, to WHAT extent - if you have immediate, near-term and long-term goals, how to put the weight on the corresponding regularizers in your objective function is a good question.